REDiscover

Tool for RNA editing discovery from NGS data.

Table of Contents

Methodology

REDiscover reports differences between transcriptome and underlying genome, these are putative RNA editing sites. To achieve that, genome and transcriptome are genotyped simultanously and basecalls are compared.

REDiscover is:

• easy-to-use - the programme will auto-detect and estimate all necessary parameters ie. strandness of your library.
• fast & lightweight, multi-core support and memory-optimised, so it can be run even on the laptop
• flexible toward many sequencing technologies and experimental designs ie. stranded and unstranded RNA-Seq, multiple genomes and/or transcriptomes are accepted as input
• reliable - the tools was tested extensively on vertebrates D. rerio

By default, REDiscover filters:

• QC failed reads
• duplicates
• reads with mapping quality (mapQ) below 15

REDiscover reports only regions fulfilling several stringency criteria:

• depth-of-coverage
• mean basecall quality

Finally, reads with basecall quality below 20 (0.01 probability of error) for given positiong are ignored.

Prerequisites

• Python 2.7+
• samtools
• pysam
• FastaIndex
• numpy
• matplotlib for plotting

All above can be easily installed with bioconda:

conda install samtools pysam FastaIndex numpy matplotlib

Usage

REDiscover input consists of aligned NGS reads (BAM) from genome(s) and transcriptomes(s). REDiscover will return a list of putative RNA editign sites and their depth of coverage and frequency across samples. Note, you can run REDiscover with RNA-seq reads alone, then you need to provide reference FastA. REDiscover will detect strandness of your library if you provide it with exon annotation (GTF or GFF). Note, mixing of stranded and unstranded libraries is not allowed!

Parameters

Most of REDiscover parameters can be adjusted manually (default values are given in square brackets []):

• General options

  -h, --help show this help message and exit -v, --verbose verbose --version show program's version number and exit -o OUT, --out OUT output file -q MAPQ, --mapq MAPQ mapping quality [3] -Q BCQ, --bcq BCQ basecall quality [20] -t THREADS, --threads THREADS number of cores to use [4]

• Reference genome (BAM or FastA)

  -d DNA, --dna DNA input DNA-Seq BAM file(s) -f FASTA, --fasta FASTA reference FASTA file

• Aligned RNA-seq reads & strandness information

  -r RNA, --rna RNA input RNA-Seq BAM file(s) -g GTF, --gtf GTF GTF/GFF for auto-detection of strandness -u, --unstranded unstranded RNAseq libraries -s, --stranded, -fr-secondstrand stranded RNAseq libraries ie. Illumina or Standard Solid -fr-firststrand stranded RNAseq libraries ie. dUTP, NSR, NNSR

• Analyse only subset of regions

  -b REGIONS, --regions REGIONS, --bed REGIONS

  BED file with regions to genotype

  -c CHRS, --chrs CHRS analyse only sublset of chromosomes [all]

• Filtering

  --minDepth MINDEPTH minimal depth of coverage [5] --minDNAfreq MINDNAFREQ min frequency for DNA base [0.99] --minAltfreq MINALTFREQ min frequency for RNA editing base [0.01] -m MAXSTRANDBIAS, --maxStrandBias MAXSTRANDBIAS max allowed strand bias [0.1] -a, --advancedFiltering enable advanced filtering (slightly more accurate, but much slower) --dbSNP dbSNP file --dist DIST distance between SNPs in cluster [300]

Test run

To run the test example, first download & unpack the test dataset:

wget http://zdglab.iimcb.gov.pl/lpryszcz/REDiscover/test.tgz
tar xpfvz test.tgz

Then execute `REDiscover.diff`:

# discover editing in RNA-seq samples (*.bam) without reference sequencing (ref.fa needed)
~/src/REDiscover/REDiscover.diff -f test/ref.fa -r test/star/*.bam -o test/editing.gz

# discover editing in RNA-seq samples (*.bam) with reference sequencing (ref*.bam needed)
~/src/REDiscover/REDiscover.diff -d test/ref*.bam -r test/star/*.bam -o test/editing.ref.gz

# if you want to ignore dbSNP sites, just add `--dbSNP snps.vcf.gz` to above commands
# or recompute only last step using `./get_enrichment.py`
## you can alter also `--minDepth`, `--minAltfreq` and many more...
~/src/REDiscover/get_enrichment.py -i test/editing.gz --dbSNP snps.vcf.gz

# violin plots for editing sites present in at least 2 samples
~/src/REDiscover/plot_violin.py -i test/editing.gz.n2.gz

# histograms for editing sites present in at least 5 samples
~/src/REDiscover/plot_hist.py -i test/editing.gz.n5.gz

For more details have a look in test directory.

Tools

Along with REDiscover, we provide a bunch of usefull tools for characterisation of RNA editing. More details about these can be find in tools directory.

FAQ

Citation

Pryszcz LP, Bochtler M, Winata CL. (In preparation) REDiscover: Robust & efficient detection of RNA editing from large NGS datasets.